[Feature] Implement pybricks.hubs.MoveHub.imu.tilt

[laurensvalk]

Is your feature request related to a problem? Please describe.
The Move Hub has had an experimental IMU implementation for a while now. There's not enough space to implement something equivalent to the other hubs.

Since V3.1 may be the last release for Move Hub, we'll want to make sure this is working the way we want. And documented accordingly.

It may be useful to see if we can implement a crude version of tilt/roll measurement, as this is really the main use case for having the IMU enabled in the first place. The user can't really do this in their code since floating point support isn't enabled, and look-up tables in user code would take too much space.

Maybe we can use libfixmath.

[dlech]

I was able to use the existing API for the R2-D2 example without problems, so I would be fine not adding/changing anything.

[laurensvalk]

That sounds good. Also, I'm not going to have time to look at this any time soon, so let's not hang V3.1 on it.

So 🚢

[laurensvalk]

It should be feasible to implement tilt using libfixmath.

As for base orientation specification, we can't use the Axis in the hub init like we do for the other hubs, but we could instead provide a Side argument to tilt to specify which side of the hub is "up" in the (0, 0) orientation.

[laurensvalk]

It should be feasible to implement tilt using libfixmath.

As for base orientation specification, we can't use the Axis in the hub init like we do for the other hubs, but we could instead provide a Side argument to tilt to specify which side of the hub is "up" in the (0, 0) orientation.

This works quite nicely: https://github.com/pybricks/pybricks-micropython/tree/movehub-tilt

But it takes up a bit too much space: 832 bytes.

This is partly due to bringing in fix16 trigonometry. This may not be worth it unless we use it for other things like pybricks/pybricks-micropython#93

[laurensvalk]

But it takes up a bit too much space: 832 bytes.

Now that libfixmath is no longer used, this implementation would be about +1000 bytes.

Here's a new variant that is about +550 bytes. Out of that 200 bytes comes from an integer implementation for atan2.

It works by interpolating a few points (green) on the positive side, and derives the full result from there (orange). It is quite close to the real result (blue).

This is accurate enough for roll computation:

The code for linear interpolation can be shared with approximations for other trigonometric functions if/when we have any.

[laurensvalk]

I'm going to give this a final push for this release:

• It makes sense for our upcoming code editor --- hub init and acceleration reads would all be the same
• Since the Move Hub has BLE broadcast, broadcasting tilt makes it a great contender for remote control.

[laurensvalk]

https://www.facebook.com/reel/1404526097073489

[laurensvalk]

Fixed by pybricks/pybricks-micropython@c42a85f and pybricks/pybricks-micropython@5e2d65d